ScienceDi ec
A ailable online a www.sciencedi ec .com
P ocedia Compu e Science 178 (2020) 27–37
1877-0509 © 2020 The Au ho s. Published by Else ie B.V.
This is an open access a icle unde he CC BY-NC-ND license (h ps://c ea i ecommons.o g/licenses/by-nc-nd/4.0)
Pee - e iew unde esponsibili y o he scien i ic commi ee o he 9 h In e na ional Young Scien is Con e ence on Compu a ional Science
10.1016/j.p ocs.2020.11.004
10.1016/j.p ocs.2020.11.004 1877-0509
© 2020 The Au ho s. Published by Else ie B.V.
This is an open access a icle unde he CC BY-NC-ND license (h ps://c ea i ecommons.o g/licenses/by-nc-nd/4.0)
Pee - e iew unde esponsibili y o he scien i ic commi ee o he 9 h In e na ional Young Scien is Con e ence on Compu a ional
Science
A ailable online a www.sciencedi ec .com
P ocedia Compu e Science 00 (2019) 000–000
www.else ie .com/loca e/p ocedia
9 h In e na ional Young Scien is Con e ence on Compu a ional Science (YSC 2020)
F amewo k o a digi al win o he Canal o Calais
Roza Ranjba a,∗, E ic Du iellaa, Lucien E iennea, Jose-Ma ia Maes eb
aIMT Lille Douai, F-59000 Lille, F ance
Uni e si y o Lille
bUni e si y o Se ille, Spain
Abs ac
The managemen o hyd og aphical sys ems is s ill mainly based on he expe ise o he manage s. Al hough his expe ise enables
he e icien managemen o hese ne wo ks unde no mal condi ions, modi ica ions due o human ac i i ies o clima e change
could lead hem o manage si ua ions ha we e no known un il now. In addi ion, ad ances in Au oma ion, Compu ing Science and
A i icial In elligence p o ide ools and me hods o assis he manage s. In pa icula , he use o ele- emo e sys ems, as SCADA,
allows he collec ion o da a and he con ol o hyd aulic de ices. Today, by bene i ing om he powe o compu e s and se e s,
he digi al wins o hyd og aphical ne wo ks can be designed. A digi al win aims o ai h ully ep oduce he dynamics o a canal.
I is used o play-back pas scena ios allowing eedback o applied managemen s a egies and as simula ions wi h p edic i e and
adap i e managemen s a egies o de e mine hei pe o mances and gi ing decision aid c i e ia o he manage s. The objec i e
o he p esen ed pape is o de ine he amewo k o a digi al win o he Canal o Calais.
c
2020 The Au ho s. Published by Else ie B.V.
This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/)
Pee - e iew unde esponsibili y o he scien i ic commi ee o he 9 h In e na ional Young Scien is Con e ence on Compu a ional
Science.
Keywo ds: Hyd og aphical ne wo ks, Wa e sys em, Digi al win, Simula ion
1. In oduc ion
In his mode n e a, he e a e s ill a huge numbe o i iga ion canals which a e managed manually despi e he la ge
wa e losses. Thus, i is c ucial o de elop he way o he canals’ managemen [18]. In addi ion o he wa e losses,
a disad an age wi h he adi ional i iga ion sys ems is he unce ain y and lack o in o ma ion wi h he a me s
ega ding he impo an en i onmen al pa ame e s like empe a u e, soil mois u e, e c. By ha ing hese pa ame e s
app op ia ely moni o ed, wa e could plays a p oduc i e ole o c ops p o i abili y. Due o hese limi a ions o he
old ashioned managemen o he wa e dis ibu ion sys ems, esea che s a e a emp ing o adop some app oaches o
i iga ion o bo h sa ing wa e esou ces and augmen ing he o e all p oduc i i y. Thus, i is now ying o eplace he
adi ional me hods o wa e dis ibu ion and supply wi h he mode n ad anced echnologies such as wi eless senso
∗Co esponding au ho .
E-mail add ess: [email p o ec ed]
1877-0509 c
2020 The Au ho s. Published by Else ie B.V.
This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/)
Pee - e iew unde esponsibili y o he scien i ic commi ee o he 9 h In e na ional Young Scien is Con e ence on Compu a ional Science.
A ailable online a www.sciencedi ec .com
P ocedia Compu e Science 00 (2019) 000–000
www.else ie .com/loca e/p ocedia
9 h In e na ional Young Scien is Con e ence on Compu a ional Science (YSC 2020)
F amewo k o a digi al win o he Canal o Calais
Roza Ranjba a,∗, E ic Du iellaa, Lucien E iennea, Jose-Ma ia Maes eb
aIMT Lille Douai, F-59000 Lille, F ance
Uni e si y o Lille
bUni e si y o Se ille, Spain
Abs ac
The managemen o hyd og aphical sys ems is s ill mainly based on he expe ise o he manage s. Al hough his expe ise enables
he e icien managemen o hese ne wo ks unde no mal condi ions, modi ica ions due o human ac i i ies o clima e change
could lead hem o manage si ua ions ha we e no known un il now. In addi ion, ad ances in Au oma ion, Compu ing Science and
A i icial In elligence p o ide ools and me hods o assis he manage s. In pa icula , he use o ele- emo e sys ems, as SCADA,
allows he collec ion o da a and he con ol o hyd aulic de ices. Today, by bene i ing om he powe o compu e s and se e s,
he digi al wins o hyd og aphical ne wo ks can be designed. A digi al win aims o ai h ully ep oduce he dynamics o a canal.
I is used o play-back pas scena ios allowing eedback o applied managemen s a egies and as simula ions wi h p edic i e and
adap i e managemen s a egies o de e mine hei pe o mances and gi ing decision aid c i e ia o he manage s. The objec i e
o he p esen ed pape is o de ine he amewo k o a digi al win o he Canal o Calais.
c
2020 The Au ho s. Published by Else ie B.V.
This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/)
Pee - e iew unde esponsibili y o he scien i ic commi ee o he 9 h In e na ional Young Scien is Con e ence on Compu a ional
Science.
Keywo ds: Hyd og aphical ne wo ks, Wa e sys em, Digi al win, Simula ion
1. In oduc ion
In his mode n e a, he e a e s ill a huge numbe o i iga ion canals which a e managed manually despi e he la ge
wa e losses. Thus, i is c ucial o de elop he way o he canals’ managemen [18]. In addi ion o he wa e losses,
a disad an age wi h he adi ional i iga ion sys ems is he unce ain y and lack o in o ma ion wi h he a me s
ega ding he impo an en i onmen al pa ame e s like empe a u e, soil mois u e, e c. By ha ing hese pa ame e s
app op ia ely moni o ed, wa e could plays a p oduc i e ole o c ops p o i abili y. Due o hese limi a ions o he
old ashioned managemen o he wa e dis ibu ion sys ems, esea che s a e a emp ing o adop some app oaches o
i iga ion o bo h sa ing wa e esou ces and augmen ing he o e all p oduc i i y. Thus, i is now ying o eplace he
adi ional me hods o wa e dis ibu ion and supply wi h he mode n ad anced echnologies such as wi eless senso
∗Co esponding au ho .
E-mail add ess: [email p o ec ed]
1877-0509 c
2020 The Au ho s. Published by Else ie B.V.
This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/)
Pee - e iew unde esponsibili y o he scien i ic commi ee o he 9 h In e na ional Young Scien is Con e ence on Compu a ional Science.
28 Roza Ranjba e al. / P ocedia Compu e Science 178 (2020) 27–37
2R. Ranjba e al. / P ocedia Compu e Science 00 (2019) 000–000
ne wo ks accomplished wi h SCADA applica ions (Supe iso y Con ol and Da a Acquisi ion) [13]. Wi h SCADA
sys ems, he manage s a e able o moni o he wa e le els and eloci ies om dis ance. They a e also able o con ol
he wa e low ia hyd aulic s uc u es which a e ully au oma ed. In his occasion, he ope a o s a e enabled o ack
he aul s and ake he necessa y main enance ac ions [3].
As he bene i s o his ield is ha , all he models can be adjus ed and in eg a ed in a compu e as a (SCADA) sys em by
employing a p og amming language such as MATLAB [2] which makes i possible o o m an ins uc ional so wa e
o con olling he i iga ion equipmen [16]. This me hod is economic and includes eliable da a o be ans e ed, in
case o necessi y, o he o he sec ions and is lexible and o ally easy o be implemen ed. I ob iously imp o es he
e iciency o he wa e dis ibu ion sys em and he supply o wa e in he equi ed domains. The senso ne wo ks can
ga he he da a om di e en ields and ans e hem o a main ope a o o he con ol cen e , o a be e decision
making [13].
The nex ad anced echnology in he abo e-men ioned domain is he use o “digi al wins”, ha p o ides dynamic op-
e a ion o simula ions. Nowadays, many wa e u ili ies a e applying digi al wins o enhance he design o he sys em
as well as op imizing he ope a ions. U ili ies a e going owa ds applying dynamic simula ion models o in eg a -
ing design and ope a ions componen s like p ocess low diag ams and SCADA sc eens wi h he piping ne wo ks o
imp o e he con ac be ween p ojec s akeholde s, pa icula ly a g oup o hem who a e p o essional and a e om
echnical backg ounds. A comple e Digi al Twin (DT) would co e whole he p ocedu e om he beginning o a aw
wa e un il i eaches he use s such ha , i obse es he whole wa e cycle. The DT s a s simula ing he sys em
pe o mance unde di e en scena ios o demands. I also allows enhancing he sys em pe o mance and p oduc i i y
ia i s abili y o op imize he decisions and lesson he p obable isks [6].
No icing he anomalies could enhance p edic i e analy ics and e o s including secu i y h ea s. DT would be use ul
in ecognizing he exac o igin o anomaly and he associa ed issues (like humans in ol ed in he loop) so ha i
could handle he physical secu i y and sys ems cybe secu i y. I is wo h no ing ha , o de eloping “sma ci ies”, DT
ope a ion p o ides a qui e accu a e in o ma ion o scale- ee ne wo ks in u ban digi al ans o ma ion. Supe ising
he digi al duplica es o wa e al es o con ol o egula e he wa e was e and wa e pollu ion a e included in he
men ioned ca ego y. Howe e , disco e ing he ull po en ial o DT will equi e mo e esea ch o imp o e he adi-
ional da a collec ion and he whole p ocedu e and i is also necessa y o apply he communica ion in e ace be ween
eal and physical wins [9]. The e a e some esea ches done wi h he ope a ion o a DT in wa e dis ibu ion sys ems
(WDS). [14] in oduced a p ima y o m o Cybe -Physical Sys em (CPS) comp ised o an Epane hyd aulic model
wi h he ma hema ical Ma lab so wa e in o de o cons uc an in elligen cybe WDS ha could implemen ce ain
al es and alloca e dis ibu ing he wa e equi ably. The wo k is amous o be a basic example o an in elligen decision
suppo sys em. La e on in 2015, [22] wo ked in inding a de ini ion o a CPS managing he quali y o wa e . Based
on his job, he CPS needs many componen s such as senso s and communica ion ne wo k, compu ing echnologies
o models, managing he se o da a and i s analysis and he p edic i e con olle s. Addi ionally, he DT mus wo k
on a hyd aulic model ha includes cohesi e and p ecise esul s o he simula ion and is well calib a ed.
Finally, he DT has o simula e hypo he ical scena ios while con aining he da a analysis and op imiza ion ools. Ma-
chine Lea ning algo i hms, like decision ees o a i icial neu al ne wo ks which p o ide modelling he complex
sub-sys ems no hidden in he hyd aulic ma hema ical model. DT could comp ise he op imiza ion algo i hms, i.e.
linea and non-linea . Figu e 1 shows a comple e unc ion o an implemen a ion o digi al wins on WDS [5]. In his
pape , he amewo k o a DT o canals is discussed ha a e dedica ed o na iga ion and e acua ion o wa e in excess
owa ds sea. To he bes knowledge o he au ho s, no DT has been designed o his ype o sys ems. The DT aims
a being used o play-back pas scena ios based on he da a collec ed om SCADA. Then, he DT will be used o es
and imp o e managemen s a egies om as simula ions.
This pape is o ganized as ollows: Sec ion 2 in oduces he managemen objec i es o he inland wa e ways, and he
amewo k o a DT design. The case-s udy o he Calais canal loca ed in he no h o F ance is p esen ed in Sec ion 3
and is used o desc ibe he unc ionali ies o he DT. Finally, Sec ion 4 summa izes he key indings and o eshadows
he possible imp o emen s in u u e esea ches.
Roza Ranjba e al. / P ocedia Compu e Science 178 (2020) 27–37 29
R. Ranjba e al. / P ocedia Compu e Science 00 (2019) 000–000 3
2. Digi al Twin amewo k o hyd og aphical ne wo ks managemen
2.1. Hyd og aphical ne wo ks managemen
Hyd og aphical sys ems a e g ea dimensional ne wo ks which a e geog aphically dis ibu ed. They a e usually
comp ised o dams, i e s, channels, e c. Hyd aulic de ices, e.g. ga es o pumps could be assigned o ou e he wa e
esou ce. Depending on he human’s needs, se e al objec i es o he hyd og aphical sys em managemen a e de ined.
Wa e is used o d inking, i iga ion, indus y, na iga ion, ec ea ional ac i i ies ( ishing, swimming, e c.); de ined as
con lic ing goals. The wa e esou ce has o be ai ly sha ed be ween usages, ollowing p io i ies ha a e de ined in
consul a ion wi h s akeholde s. Mo eo e , o each objec i e, e.g. i iga ion, he wa e has o be con olled wi h e i-
ciency, a oiding lack and excess o esou ce o e la ge ime ho izons ha can be seasonal, o mul i-annual in se e al
pa s o he wo ld.
The wa e esou ce managemen is usually pe o med based on he manage ’s knowledge. This knowledge can be
syn hesized by expe ules and hen au oma ically eused o ace some known e en s. Expe ules lead o an e icien
managemen o hyd og aphical ne wo ks. In addi ion, he manage s o hyd og aphical sys ems ook he oppo uni y
o in oduce new con ol algo i hms [23] o op imiza ion app oaches [7] ha we e designed in he pas yea s. Howe e
in he con ex o clima e change, popula ion g owing linked o he u baniza ion and he modi ica ion o usages; unex-
pec ed si ua ions can occu showing he limi s o expe ules, he designed con ol o he op imiza ion algo i hms.
Acco ding o he nowadays de iciency in he wa e esou ces (d ough pe iods) o he ex eme ain e en s, he man-
agemen o hyd og aphical sys ems a e eally impo an . The al eady p oposed echniques ha a e dedica ed o he
managemen o hyd og aphical ne wo ks can bene i o ad ances in Au oma ion, Compu ing Science and A i icial
In elligence. Based on he powe o compu e s and se e s, i could be possible o design Digi al Twins o hyd og aph-
ical ne wo ks wi h he aims i) o ep oduce he dynamics o he eal sys ems, ii) o iden i y unknown inpu s/ou pu s,
iii) o play-back pas scena ios p o iding a eedback on he applied managemen s a egies, i ) o es and op imize
new p edic i e con ol s a egies. The amewo k o a DT is desc ibed in he nex subsec ion.
2.2. Digi al Twin amewo k
The amewo k o he DT o a hyd og aphical ne wo ks is based on a WDS s uc u e on which a hyd ological
model can be associa ed (see Figu e 1). The hyd ological model aims a p edic ing he uno acco ding o he ain by
conside ing se e al p edic i e ho izons. This in o ma ion is e y use ul o educe unce ain ies on inpu s and o ha o
p edic i e con ol algo i hms. In case ha no hyd ological model is a ailable, he hyd aulic model and collec ed da a
can be used o es ima e he unknown inpu /ou pu . I is ,howe e , necessa y o ha e accu a e models o he hyd aulic
de ices and accu a e measu es (discha ges, o le els). In e se modelling app oaches a e used o es ima e he unknown
inpu /ou pu o he pas e en s.
The DT is hen connec ed o a so wa e such as Ma lab o o a compu e code and p og ams ha implemen he expe
ules o he p edic i e and adap i e con ol s a egies. The objec i e is o o e a sui able en i onmen o designing
he new con ol echniques and managemen s a egies ha can be uned and es ed using he DT.
The design o he DT equi es some s eps: a) hyd aulic model o he hyd og aphical sys ems, b) iden i ica ion o
he dynamics o he con olled de ices and c) calib a ion o he hyd aulic model.
Hyd aulic models o he open- low channels a e based on Sain Venan equa ions ha accu a ely ep oduce he gen-
e a ion and p opaga ion o he wa es wi h a a iable delay and a enua ion. The e a e some solu ions, wi h a non-
exhaus i e lis , SIC21, Hyd a2, Mike113, SWMM4, HEC RAS5. Among hese solu ions, HEC RAS p esen s he main
ad an age o be ee and SIC2is dedica ed o he con ol s a egies design. Mos o hese solu ions can be linked o
1h p://sic.g-eau.ne /?lang=en
2h p://hyd a-so wa e.ne /
3h ps://www.mikepowe edbydhi.com/p oduc s/mike-11
4h ps://www.epa.go /wa e - esea ch/s o m-wa e -managemen -model-swmm
5h ps://www.hec.usace.a my.mil/so wa e/hec- as/
30 Roza Ranjba e al. / P ocedia Compu e Science 178 (2020) 27–37
4R. Ranjba e al. / P ocedia Compu e Science 00 (2019) 000–000
Figu e 1. F amewo k o he DT o hyd og aphical ne wo ks.
GIS (Geog aphic In o ma ion Sys em), e.g. QGIS 6is e y in e es ing in case o he opog aphy o he canals; so ha
i is di ec ly accessible.
The e is an exhaus i e li e a u e on hyd aulic de ices while hey a e cha ac e ized by he nonlinea dynamics. Some
o he so wa e solu ions could in eg a e models o hyd aulic de ices (wei , ga e, e c.), howe e , o hyd aulic de ices
p esen ing mo e complex dynamics, Machine Lea ning echniques a e a ailable o iden i y black-box o g ey-box
models based on he measu emen . These echniques, e.g. Suppo Vec o Machine (SVM), a e sui able o nonlinea
and hyb id dynamics. Al hough, he echniques equi e a ich and exhaus i e da abase.
Finally, when i comes o he calib a ion o he hyd aulic models, he ic ion coe icien is used in gene al. He e again,
a ailable da a a e equi ed.
All he abo e-men ioned s eps aim a ep oducing he dynamics o he hyd og aphical ne wo ks.
2.3. Digi al Twin usage
One o he ou objec i es o he DT lis ed abo e consis s an accu a e ep oduce o he eal sys ems’ dynamics.
As i has been explained in he p e ious subsec ion, once a calib a ed model is a ailable, i is possible o iden i y
he unknown inpu s/ou pu s o a hyd og aphical sys em. By conside ing eal da a, le els and con ols ha ha e been
sen o he hyd aulic de ices du ing a pe iod o ime and a known ini ial condi ion, he wa e olume balance is com-
pu ed. Thus, he missing lows (posi i e o nega i e) can be easily de e mined by he conside ed pe iod o ime (o he
echniques like Kalman il e [15], obse e [1] o Mo ing Ho izon Es ima ion (MHE) [19] app oaches could also be
in es iga ed). Once he unknown inpu s/ou pu s es ima ed, he pas scena ios can be play-backed and a eedback on
applied managemen s a egies can be pe o med as i is p oposed in [8]. The pe o mance o he applied s a egies
a e de e mined p o iding use ul in o ma ion o he manage s. Mo eo e , ano he con ol o op imiza ion echniques
can be applied using he same scena io, i.e. he same unknown inpu s/ou pu s. In his case, he con ols sen o he
hyd aulic de ices should no be he same, and new s a es ha would ha e had he hyd aulic sys ems acco ding o
hese new se poin s a e compu ed hanks o he DT. The pe o mance o he applied managemen s a egies and o he
con ol/op imiza ion echniques could be compa ed wi h each o he .
The con ol o op imiza ion algo i hms ha e o be designed be o e conside ing eal pas scena ios which he DT o -
e s his possibili y. The algo i hms which a e gene ally based on models gained om physics o assump ion o he
linea i y o he canal dynamics [11,10,4,12,21,20,17,19] could be uned and es ed.
6h ps://www.qgis.o g/en/si e/
Roza Ranjba e al. / P ocedia Compu e Science 178 (2020) 27–37 31
R. Ranjba e al. / P ocedia Compu e Science 00 (2019) 000–000 5
Figu e 2. Map o he wa e sheds in no h o F ance and o he Wa e ingues Te i o y ( ed iangle).
3. DT o he canal o Calais
The objec i e o his sec ion is o illus a e he s eps o he DT by conside ing he canal o Calais; loca ed in
he no h o F ance. The pa icula i y and he managemen objec i es o he canal o Calais a e i s ly desc ibed.
This canal is modelled wi h he so wa e SIC2by conside ing a e age physical cha ac e is ics. Then, he me hod o
ob ain a dynamical model o he hyd aulic de ices is p esen ed. Finally, he unknown inpu s o he canal o Calais a e
es ima ed based on an iden i ica ion echnique.
3.1. Desc ip ion
Loca ed in he no h o F ance, he Calais Canal is in a e i o y called Wa e ingues. The Wa e ingues is he
ex ension o a polde called “Yse ” and is on he way o h ee owns named Sain -Ome , Calais and Dunke que.
Ac ually, he e i o y builds a iangle a ea o 100000 hec a es (see Figu e 2). I co ela es wi h he Aa del a; ha
used o be a la ge we land in he pas . The e i o y is si ua ed below he sea le el (wi hin he class o lowland) and
he e a e di e en wa e sys ems such as le ees, ga es o sea and wa e pumps wi hin he a ea.
The wa e o he canal is s o ed by he wa e gangs, which a e no mally di ches o d ainages o dehyd a e he lowlands
[8]. The wa e gangs se up a gian meshed ne wo k which leng h is mo e han 1500 km. The su plus wa e inside
he wa e gangs could be pumped o he na iga ion canals loca ed along he Calais canal. On he o he hand, in he
no he n pa o he e i o y ( a om he main e i o y), he e is a zone called “Les Mo¨
e es”, in which he wa e
was leading o a sc ew highe han he sea le el 7.
The e is a p incipal each in he Calais canal ha is supplied by h ee sub-canals named Aud uicq canal, A d es canal
and Guines canal (see Figu e 3). The lock o Hennuin is loca ed in he ups eam which is basically used o he ma e
7h p://www. loodcom.eu/pa ne s-si es/si es/ ance-s -ome /index.h ml

32 Roza Ranjba e al. / P ocedia Compu e Science 178 (2020) 27–37
6R. Ranjba e al. / P ocedia Compu e Science 00 (2019) 000–000
Lock o
Hennuin
Mowe
{0; 0.35}
SP 7
{0; 0.2}
Rebus
{0; 0.4}
Pon Neu
{0; 0.5}
Nlle Eglise
{0; 0.45}
SP 6
{0; 0.2}
Gde Mee s ae en
{0; 0.65}
GdVin il
{0; 0.8}
Canchoise sud
{0; 0.18}
Ancne
Canchoise
{0; 0.2}
3 co ne s
{0; 1.3}
A aques
{0; 0.45}
No ke que
{0; 0.75}
Sud Bou illez
{0; 0.2}
Cana de ie
{0; 0.2}
Lac d’A d es
{0; 0.28}Balinghem
{0; 1}
Po ez
{0; 1}
Ga es
Calais
{0; 8}
Ba elle ie
{0; 4}
Pumping s a ion Lock
Ga e
Za aq
Zcalais
Le el me e
Figu e 3. Schema ic iew o he Calais canal.
o na iga ion. In he downs eam o he canal, he e a e sea ou le ga es con ibu ed by wo pumps in Calais and wo
pumps in Ba elle ie. The own capaci y o he pumps o Calais is 4m3/sand ha o Ba elle ie is 2m3/s. The e a e
wo le el-me e s which make i possible o measu e he le el in Les A aques and in Calais, de ined Za aq and Zcalais,
espec i ely.
The Calais canal is equipped wi h 18 Pumping S a ions (PS), e.g. PS Mowe o PS Cana de ie , ha a e loca ed
along he each. The PSs a e ope a ed by a me s wi hin hei own schedule. When he PS is o , he discha ge is
equal o ze o and when i is on, he a e age discha ge is known and is indica ed in Figu e 3in b acke s, e.g. PS
Mowe wi h a discha ge equal o 0.35m3/s o pump is on. When all he PSs a e ope a ed, he incoming low is equal
o 8.46m3/s. The maximum lows o he h ee seconda y canals a e espec i ely QAud uicq =3m3/s,QA d es =1m3/s
and QGuines =0.2m3/s, e en i he manage s suspec much highe lows du ing ex eme ain all e en s. Tha means
ha he maximum inpu low is equal o 13.06m3/s, om PS and om uno .
The wo main objec i es o he managemen o he Calais canal a e na iga ion and lood a oiding. Fo he na iga ion
pu pose, he le el in he canal has o be kep close o he No mal Na iga ion Le el (NNL) and inside he na iga ion
ec angle: an in e al ha is de ined wi h wo o he le els; High Na iga ion Le el (HNL) and Low Na iga ion Le el
(LNL). Usually, HNL =NNL+30cm and LNL =NNL−30cm. To a oid looding, he managemen includes ejec ing
wa e in excess o sea hanks o he ou le ga es and he pumps in Calais and Ba elle ie, acco ding o he sea ide.
The ou le ga es can be opened only du ing he low ides. Fo economical easons, he pumps ha e o be ope a ed as
li le as possible so ha he d awdown zone o he canal is used. The le el can inc ease closely o he HNL du ing a
high ide wai ing o he nex low ide when he ou le ga es can be ope a ed. Du ing he low ide, he ou le ga es a e
opened leading o a le el o he canal close o he LNL, o e ing he bigges s o age capaci y o he nex high ide.
Then, he le el can oscilla e a ound he NNL, limi ing he use o pumps. Howe e when he uncon olled inpu lows
a e oo impo an , he pumps ha e o be ope a ed wi h he aim o a oiding o a leas limi ing he lood.
A e modelling he Calais canal in SIC2, as a i s s udy, an a e age p o ile o he Calais canal is conside ed. The
leng h, wid h and dep h o he canal a e L=26.72km,W=20mand D=2.2m, espec i ely. The a e age low
is Q0=1m3/sand he Manning’s oughness coe icien is n =0.035m1/3/s. The so wa e SIC2is easily linked
wi h Ma lab [8]. Some simula ions a e pe o med by using s ep a he ups eam o he canal. The ob ained delays
and a enua ion o he hyd og aphs a e close o hose obse ed on he eal sys em. These es s aim a alida ing he
pe o mance o he selec ed so wa e. I is hen necessa y o de e mine he model o he ou le ga es dynamics.
Roza Ranjba e al. / P ocedia Compu e Science 178 (2020) 27–37 33
R. Ranjba e al. / P ocedia Compu e Science 00 (2019) 000–000 7
3.2. Es ima ion o he ou le ga e dynamics
The e a e wo ou le ga es, G1and G2. They a e in pa allel, bu he ga e G2is only opened when he ga e G1is
comple ely opened. The a e age discha ge o G1depends on he opening o he ga e and on he cycle o ide ha
is composed o a sp ing ide and a dead ide. The ype o ides depends on he ide’s coe icien ha can be easily
o ecas ed; coe icien s highe han 80 co espond o sp ing ides while coe icien s smalle han 45 co espond o
dead ides. Ano he ype o ide has o be conside ed o coe icien s be ween 45 and 80; he a e age ide.
Discha ges o ew openings ha e been es ima ed by he manage (see Table 1) acco ding o he ype o ide. I was
necessa y o de e mine unc ions o es ima e he a e age discha ge o all he ga e openings om 0dm o 25dm. These
unc ions a e iden i ied acco ding o he a ailable da a by conside ing he ma hema ical unc ion o ela ion 1. The
pa ame e s a1,b1and c1ha e o be de e mined acco ding o he a ailable da a o each ype o ide. The nonlinea
eg ession nlin i o Ma lab is used o de e mine he alue o he pa ame e s which a e gi en in Table 2and he plo o
he unc ions a e shown in Figu e 4. This Figu e shows a good es ima ion o he discha ge unc ions.
Table 1. Discha ge QG1[m3/s] s ga e OG1opening [dm].
OG10124681012151720
Qsp ing ide
G10 1.9 2.7 5.2 7.6 9.4 11.1 12.3 13.9 14.3 14.7
Qa e age ide
G10 1.7 2.4 4.8 7 8.7 10 11.1 12.5 12.9 13.4
Qdead ide
G10 1.4 2.2 4.3 6.3 7.9 8.9 9.8 11.1 11.5 12
(OG1)=a1(1−exp(−b1.OG1
c1)) (1)
Table 2. Es ima ed pa ame e s o he nonlinea unc ion linked he discha ge QG1[m3/s] o he ga e opening OG1[dm].
Pa ame e s a1b1c1
Sp ing ide 16.7366 0.0854 1.0986
A e age ide 14.3988 0.0907 1.1206
Dead ide 13.2299 0.0891 1.1025
Conside ing he second ga e G2, when OG1=25dm and G2is open, he a e age discha ges a e equal o 15.25m3/s
du ing sp ing ide, 13.9m3/sdu ing a e age ide and 12.55m3/sdu ing dead ide.
3.3. Es ima ion o he unknown inpu s
The p oposed app oach o es ima ing he unknown inpu s is based on he simula ed model and eal da a. This i s
app oach aims a being simple and easy o use. I consis s de e mining he di e ence o olume in he canal acco ding
o ela ion 2.
∆V(k)=∆Z(k).SCalais (2)
wi h SCalais he a ea o he canal such as SCalais =L.Wand ∆Z(k)=ZCalais(k)−ˆ
ZCalais(k), whe e ZCalais(k)is he le el
measu ed in Calais and ˆ
ZCalais(k)is he es ima ed le el om he so wa e SIC2.
The di e ence o olume ∆V(k)is a e aged on a ime window ∆Tleading o µ∆V|∆T. The di e ence o discha ges
be ween wo pe iods o ime is gi en by ela ion 3. These alues co espond o he unknown inpu s on he Calais
canal.
∆Q=µ∆T+1
∆V−µ∆T
∆V
∆T
(3)
34 Roza Ranjba e al. / P ocedia Compu e Science 178 (2020) 27–37
8R. Ranjba e al. / P ocedia Compu e Science 00 (2019) 000–000
Figu e 4. Es ima ed discha ge QG1[m3/s] unc ion o ga e opening OG1[dm]acco ding o he ype o ide, wi h he poin s gi en by manage s (s a ),
and he es ima ed poin s (do ).
Figu e 5. Coe icien s o he ide o No embe 2019.
3.4. E en o No embe 2019
In he beginning o No embe 2019, an ex eme ain led o ou pe iods o lood ( he 5 h and 6 h o No embe (see
Figu e 8.c)). This pe iod o 13 days is selec ed o illus a e he p oposed s ep. The coe icien s o he ide is gi en
in Figu e 5, p esen ing pe iods o sp ing ide he 1s , 11 h and 12 h, a e age ide o he 2nd,9
h and 10 h, dead ide
o he wise. The ac i a ion o PSs is p esen ed in Figu e 6 o PSs ha a e loca ed close o he PS 3 Co ne s. Due o a
lack o space, o he PSs a e no p esen ed. The ope a ions o he pumps in Calais a e shown in Figu e 7. The pumps
in Ba elle ie a e con inuously on om No embe 5 h o 9 h.
Based on he le el in Calais and he simula ed le el om he so wa e SIC2, on he unc ions ha a e used o
es ima e he a e age discha ge a he ou le ga es (see Figu e 8.a), he unknown inpu discha ges a e es ima ed by
conside ing a pe iod o ime ∆T=6hou s (see Figu e 8.b). These es ima ed unknown discha ges allow o ob ain a
simula ion o eal scena io using he DT (see Figu e 8.c). The sea le el is depic ed in Figu e 8.d. I can be obse ed
ha he o al discha ge om he seconda y canals is highe han wha was expec ed and he le el om he DT is close
o he eal one.
Howe e , some big di e ences could be obse ed a he end o he low ide he 3 d ,8
h and 9 h o No embe . These
di e ences o igina e om he cons an a e age discha ges o he ou le ga es. In eali y, he discha ge is he mos when
he ga es a e open and a e wa ds, he e would be a loss o discha ge a he end o he low ide. This dynamics is no
Roza Ranjba e al. / P ocedia Compu e Science 178 (2020) 27–37 35
R. Ranjba e al. / P ocedia Compu e Science 00 (2019) 000–000 9
Figu e 6. S a e o PS loca ed close o he PS 3 Co ne s in No embe 2019.
Figu e 7. S a e o he pumps in Calais in No embe 2019.
well modelled and in u u e wo ks, a mo e accu a e dynamics o he ou le ga es will be p oposed. Once he unknown
discha ges a e es ima ed, i would be possible o play-back he scena io and gi e a eedback on he managemen
s a egies. I seems ha ope a ing he wo pumps in Calais could lead o a limi a ion o he lood pe iods.
4. Conclusion
In his pape , a amewo k o a digi al win is p oposed o ep oduce he dynamics o he Calais canal and o
implemen he ad anced con ol on he hyd aulic de ices o he canal. The canal is simula ed by he so wa e SIC2.
The s a egy behind he new managemen o he canal ollowed he goals o na iga ion and p ohibi ion/limi a ion o
lood e en s. Fo achie ing hese wo objec i es, he wa e le el and he ou le ga es o he canal should ha e been
con olled. Thus, he analysis ha e done on bo h he ou le ga es dynamics and he unknown inpu s. A e wa ds, he
simula ed le el o he wa e and he discha ge a he ou le ga es ha e been inspec ed. Al hough, he e we e some
di e ences in he simula ed igu es due o he unknown discha ges, i is s ill possible ha a e ecognizing hem, by
doing a play-back o he scena ios, he e would be an enhance in he managemen s a egies. Addi ionally, ope a ing
he pumps showed a limi a ion in he lood pe iods. Fu u e wo ks will consis in imp o ing he models o he ou le